This invention is related in general to video conferencing systems and more specifically to devices, systems, architectures, and methods for using media notifications to coordinate switching between video in a distributed arrangement.
In conventional video conferencing, typically all of the video screens are active when the conference is in session in a “continuous presence” mode of operation. However, newer methods don't send a conference participant's media representation into the network unless they are, for example, the loudest talking speaker. Such approaches aim to save bandwidth across the network because not all of the participants need to be seen in the conference. In fact, only video streams to be presented to other participants need be transmitted onto the network.
Referring now to FIG. 1, a block diagram illustrating a conventional centralized video conferencing control architecture is indicated by the general reference character 100. Location A may include endpoint (EP) 102A, EP 104A, and input/output (IO) 106A. Similarly, location B may include EP 102B, EP 104B, and IO 106B. Endpoints may represent particular conference rooms, for example. Communication channel 114 may connect location A with location C via IO 106C, which can also interface to EP 102C and EP 104C. Similarly, communication channel 116 may connect location A with location C via IO 108C, which can also interface to EP 110C. Multipoint control unit (MCU) 112 may interface with IO 106C via channel 118 and with IO 108C via channel 120.
Because IO systems 106A, 106B, 106C, and 108C are typically not fully dedicated or sufficiently adapted to accommodate video conferencing, system performance may suffer. Further, relatively large bandwidth may be required for communication channels 114 and 116. Even more bandwidth may be required for channels 118 and 120 due to the centralized MCU 112 configuration. In addition to possible bandwidth bottlenecks in this conventional approach, high speed and “smooth” video stream switching from a previous speaker to a new active speaker is not optimized.